FE model validation using experiments
The present three-dimensional FE model was validated experimentally for the hip implant alone (Construct A) at loads of 700 and 2000 N, showing reasonable agreement (Table 1). The Modular implant showed an overall average difference for all locations of 3.0% between FE and experimental values. The Monoblock device showed an average difference of 3.2% for the proximal Locations 1, 2, 5, and 6; however, the average difference was 53.4% for the distal Locations 3, 4, 7, and 8. This can be attributed to the medial slip of the femoral head under the load application plate that was visually observed (but not measured) during experiments, since motion was not constrained in this direction, causing exaggerated distal strains in the Monoblock device. This was not replicated in the FE model. In addition, it is likely that the lower stiffness Monoblock stem underwent medial slip more so than the higher stiffness Modular stem.
Modular versus Monoblock hip implants
For Construct A (Table 1), FE and experimental strains showed that the Modular implant was about 3 times mechanically stiffer than the Monoblock, and the stiffness measurements showed the Modular device to be 4.5 times stiffer. This was most likely due to the difference in geometry of the devices with respect to their cross-sectional diameters in the transverse plane for the distal two-thirds of their lengths, namely, Modular (range, 16.5 to 20.2 mm) and Monoblock (range, 15.4 to 16.3 mm) (Figure 1). The material properties of the stems were virtually identical, since they were both made from titanium-based alloys. In the FE model, in fact, both implants had identical material properties assigned to them. Higher implant stiffness may create an implant-femur construct with an increased stiffness that provides enhanced mechanical stability in the immediate post-operative situation, but it may create eventual femur stress shielding and bone loss . Conversely, lower device stiffness may result in increased load transfer to the host femur, subsequently stimulating bone ingrowth around the implant, minimizing bone loss due to resorption, and improving mechanical stability in the long-term [23–27].
For Construct B (Figure 6a and 6c), FE results showed that there was better overall load transfer to the whole femur and, thus, less load absorption by the Monoblock stem itself (0-75 MPa, excluding peaks) compared to the Modular implant (0-85 MPa, excluding peaks). Therefore, the Monoblock prosthesis might facilitate greater load transfer to the femur, potentially promoting increased bone ingrowth around the implant and reducing bone remodeling and resorption [23–27]. Conversely, the Modular device carried a greater amount of load, making it is less conducive to biomimetic performance in the patient femur. Moreover, the FE results identified a stress concentration on the stem at the modular junction of the Modular implant (Figure 7). This could make it susceptible to fracture clinically, in which the load levels may be elevated compared to that currently used, especially during higher impact activities which patients might choose to perform .
Comparison of strains to prior studies
No earlier studies have examined the mechanical characteristics of the present Modular and Monoblock implants. However, comparison of current results to prior experiments on primary hip prostheses may be instructive. Waide and colleagues  found that experimental compressive microstrains on the medial side peaked at 756 for Muller-Curved and 765 for Lubinus SPII hip prostheses cemented into synthetic femurs. Akay and Aslan  measured experimental peak microstrains for hip implants cemented into synthetic femurs, held in 20 deg of adduction, and compressed with a 3000 N axial force. They obtained microstrains of almost 4000 (medial) and 2500 (lateral) for a carbon-based composite implant and about 1200 (medial) and 900 (lateral) for a titanium alloy prosthesis. These are similar to experimental strains achieved at present for Construct A (Table 1) and for FE analysis strains for Construct B (Figure 5) at 2000 N. Even so, proper inter-study comparison is difficult because of the variety of implant materials, implant geometries, and experimental conditions used in the literature, resulting in a broad range of strain levels achieved on implant surfaces.
Prior clinical results showed no problems at the Modular junction, though follow-up was limited to 2 and 5 years [10, 11]. Other clinical results showed that the junction is a site for potential implant fracture, often being associated with proximal bone loss causing implant loosening and increased surface stresses [29, 30]. This agrees with maximum stem strains noted presently for the Modular device at its modular junction (i.e., at the proximal and distal interface) (Figure 5). This may lead to premature implant failure, especially at higher loads than those examined at present . The Modular device transferred higher loads to the distal half of the femur bone itself (0-7.5 MPa, including peaks) than the Monoblock device was able to achieve (0-6.9 MPa, including peaks) (Figure 6b and 6d). As confirmed clinically [31, 32], increased load transfer to the distal femur could cause bone densification around the distal implant tip and possible femur fracture. Conversely, the single-piece Monoblock implant had limited strain concentrations towards the proximal neck with increasing strains cascading down towards its distal end. This was punctuated by a peak stress at the distal-most medial point near the cement chamber. While this did not replicate a clinical scenario, it did suggest that the Monoblock device offered more anatomical loading than the Modular implant. Accordingly, the Monoblock prosthesis may have replicated the natural femur better, promoting superior load transfer.
Firstly, only axial compression tests were assessed. The hip device would clinically be exposed to dynamic forces exceeding the 700 and 2000 N used presently due to muscle action and activities of daily living [21, 33]. Given identical host bone quality, muscle activity, and soft tissue constraint, each of which have an influence on the dynamic forces and moments experienced by the bone-implant construct, it is anticipated that the relative performance reported of the Modular versus the Monoblock hip prostheses would be similar in a "real world" clinical scenario. However, the absolute values of stress and strain obtained at present may not be reflective of in vivo results. A direct biomechanical comparison between the stems in vivo would be required to demonstrate these postulations conclusively.
Secondly, the FE model used to assess Construct B assumed that the synthetic femur had linear isotropic mechanical properties. This simplified the analysis significantly. However, nonlinearity, anisotropy, and viscoelasticity may influence bulk mechanical behavior of the bones. Even so, comparison of FE analysis, synthetic femurs, and human cadaveric femurs previously performed by some of the authors  showed that linear behavior is a good approximation of the femurs in axial compression and torsion. Moreover, these assumptions have been used in the literature previously for synthetic femur FE modeling [14–16, 26].
Thirdly, FE analysis had limited success in assessing the experimental performance at the distal end of the Monoblock device for Construct A. This was likely due to the visual observation that the implant shifted medially under the load application plate, since motion was not constrained in that direction, thereby exaggerating distal strains. This was not replicated in the FE analysis. Moreover, it is possible that the lower stiffness Monoblock stem underwent such medial slip more so than the higher stiffness Modular stem. Nonetheless, the FE model was able to replicate all Modular strains and proximal Monoblock strains to within about 3% of experimental data.
Fourthly, the FE model mimicked long-term bone ongrowth around the stem and mechanical stability by employing bonded contact at all implant-bone, implant-cement, and implant-implant interfaces, i.e., bone absorption/apposition was neglected. However, in actuality, ongoing bone remodeling over the life span of the implant will alter implant-femur biomechanics. The current FE model, in effect, only assessed the biomechanical situation at one time point during the service life of the hip stem.
Fifthly, equivalent Von Mises strains from FE analysis were used for comparison to experimental values from uniaxial linear strain gages. It could be suggested that a more appropriate technique would have been to compare only FE principal strains with linear strain gage results. However, it must be noted that the current approach has been used by some of the present authors and others in prior studies of bone-implant biomechanics [15, 16]. Moreover, the strain gages are physically large, so that they actually detect the average surface strain over a given area, which is different than the principal strain at a specific point. Furthermore, comparing equivalent Von Mises strains from FE with experimental values may have been especially appropriate for slightly curved surfaces, i.e. Locations 1, 2, and 5, since the slightly curved uniaxial strain gages were essentially providing an average reading of strain along the length of an arc. Ideally, however, rosette-type strain gage setups may be able to provide more exact experimental readings for comparison to Von Mises strains from FE analysis, especially when circumferential strains are also of interest [19, 20].
Finally, this study combined into an FE implant-femur model (Construct B) separate FE models for the hip implants alone presently and the synthetic femur from prior investigations [14–16]. Although individual implant and femur FE models have been validated, future work may ideally include experimentally testing an actual implant-femur construct. The reason the present authors performed separate experimental validations of the FE models for the stems alone and the femur alone was a pragmatic one. Specifically, physically inserting hip stems into the synthetic femur canal would have made strain gage measurements along the surface of the hip stem itself extremely difficult experimentally because of the very fragile nature of strain gages (i.e. they easily peel or shear off) and the confined space inside the canal (i.e. there is little room to accommodate strain gage wiring).
In spite of these drawbacks, the main intention of this study was only to provide a preliminary assessment of the biomechanics of the Modular and Monoblock stems upon which more comprehensive future work may be built.